No More Bradley Effect?
Some Harvard egghead has concluded the so-called "Bradley/Wilder Effect" -- broadly defined as polls showing falsely high numbers for a minority candidate versus their actual performance -- no longer exists. Whomever you support, it's always good news that people feel comfortable expressing their true preferences without fears of appearing racist.
posted by QueensGuy @ 1:21 PM
12 Comments:
I question 1) if the effect ever existed--polls can be wrong for lots of reasons and 2) if it did exist, if there's any way of knowing how it would work in the present election.
Did you have a chance to read the linked article? I think it addresses your first objection pretty comprehensively. Another interesting finding was that there's no evidence of it ever having happened for gender -- the so-called Ferrarro Effect -- and no evidence that the Bradley Effect persisted past the early 90s.
No matter how rigorous your numbers are, there are alternate explanations that correlate very well with the Bradley effect.
Ok, I'll bite. Beside poll sampling errors, what is one such alternative explanation?
The Bradley Effect means people are saying one thing (they'll vote for a guy) and doing another (don't vote for the guy) for a particular reason (his race). I'm simply saying perhaps they said they'd vote for the guy, and/or ended up not voting for the guy, for some other reason that correlates with the reason, but isn't the reason--they're the kind of people who'd say one thing and do another.
I'm not sure I follow. I think you're saying that black candidates have repeatedly polled higher among liars, for an unknown reason that correlates with lying, but that is something other than the candidate's race. Yes?
We know there are liars. Now the question is why. Just because there seems to be a correlation with the race of the candidates in a few cases doesn't mean there can't be another correlation.
Now hold on -- you're trying to change the terms of the debate. First you said "no matter how rigorous your numbers are," but now you're trying to back off from that with "in a few cases." Let's stick with the former. Unless I've misunderstood your position, it seems that for you to accept the Bradley Effect as being a result of the race of the candidate, I'd need to show not only a statistically meaningful correlation between lying in polls and the race of the candidate, but that there is no other correlation with any other factor. In other words, I'd need to prove a negative. Please tell me whether I've misunderstood, because that position bears a striking resemblance to the racism-denial alleged by the anonymous commenter in another thread.
How did this become about racism denial, especially since even if you accept the Bradley effect, it doesn't have to equate to racism. I'm simply claiming, even if you believe the data is rigorous (just how many elections would you need for that, no matter how many numbers you have for each case?), there could be a deeper factor causing both the voting totals and the apparent Bradley effect. As an example, let's say someone claims women tend to vote a certain way, and backs it up with tons of evidence, far more than the Bradley Effect study. Well, is it because they're women? Perhaps, but perhaps the real reason is because people with less money vote a certain way, and also, people will less money are more likely to be women. That's all I'm saying.
The Bradley Effect doesn't equate to racism, it equates to politican correctness.
Re racism denial, the anonymous commenter's point (as I recall it, anyway) was that some folks here tend to discount racism as a real, persistent phenomenon. Setting an unreasonably high bar to prove that attitudes toward race (though not necessarily racism, I agree) were responsible for behavior is uncomfortably close to that for me. I'm imagining an employment discrimination jury case where a juror took a similar position -- "hey, forget their statistical expert's unrefuted evidence that blacks were promoted less than whites; that may just be a result of some underlying correlating factor that we're not aware of."
All I'm saying is that if it quacks like a duck, reacting by saying "I'm sure there are a multitude of species with which I'm unfamiliar that may sound like that," tends to sound a bit unreasonable.
Anon., thanks for making that point. I suppose that's what I (poorly) tried to say in the original post -- people not feeling the need to give a politically correct answer any more is a good thing.
Post a Comment
<< Home